fix(eval): resolve cross "topmost" via the order matrix, not array position #25

New issue

Closed

opened 2026-07-02 14:03:37 +00:00 by toph · 0 comments

toph commented

2026-07-02 14:03:37 +00:00

Owner

The spec resolves a cross intersection on stacked layers to "the material point on the topmost layer". The implementation takes the last preimage in face-array iteration order (eval.ml:63) — but fold_state.ml:5 states that array index carries no z-meaning and all stacking lives in order. This only works today because of the deliberately preserved legacy append order plus flip reversing the array.

The moment a fold inserts a face between existing layers (pocket slice), cross silently returns the wrong material point. The covering faces pairwise overlap at the point, so their restriction of order is total and "topmost" is well-defined — consult the order matrix instead.

Must land before the pocket slice.

The spec resolves a `cross` intersection on stacked layers to "the material point on the topmost layer". The implementation takes the *last* preimage in face-array iteration order (`eval.ml:63`) — but `fold_state.ml:5` states that array index carries no z-meaning and all stacking lives in `order`. This only works today because of the deliberately preserved legacy append order plus `flip` reversing the array. The moment a fold inserts a face between existing layers (pocket slice), `cross` silently returns the wrong material point. The covering faces pairwise overlap at the point, so their restriction of `order` is total and "topmost" is well-defined — consult the order matrix instead. Must land before the pocket slice.

toph referenced this issue from a commit

2026-07-02 20:55:44 +00:00

fix(fold): resolve cross-topmost via order matrix and emit M/V for folded precreases (#25, #27)

toph referenced this issue

2026-07-02 20:56:18 +00:00

fix(fold): cross-topmost via order matrix; M/V for folded precreases #34

toph referenced this issue from a commit

2026-07-02 22:07:04 +00:00

fix(fold): cross-topmost via order matrix; M/V for folded precreases (#34)

toph closed this issue

2026-07-02 22:07:05 +00:00